Amazon reportedly buys and destroys rare books to train AI models
The company reportedly purchases physical books, removes their bindings and scans the pages as AI developers seek higher quality training data.
Amazon is reportedly purchasing large quantities of rare books, scanning their contents for AI training data and destroying the physical copies in the process, according to 404 Media.
The publication tracked one shipment after placing an Apple AirTag inside a rare book included in an order of roughly 1,000 titles. The shipment ultimately arrived at an Amazon facility in Las Vegas that workers said is dedicated to scanning books.
Amazon employees described a process in which books are received, catalogued and then have their bindings removed so pages can be scanned more efficiently. The destructive scanning process allows companies to digitize large quantities of text at lower cost.
An Amazon spokesperson confirmed that the company purchases books through commercial channels to help develop and improve its products and services, but did not provide additional details about how the resulting data is used.
AI companies have increasingly sought books as training material because professionally edited texts can provide higher quality writing data than much of the internet.
AI, tech, and the markets they move—in one daily briefing.
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
Books published before the widespread adoption of generative AI are also valuable because they are less likely to contain AI generated material.
Amazon is not the only AI company reportedly using destructive book scanning. Anthropic previously operated a large scale effort known internally as Project Panama to purchase, scan and destroy physical books for training data.
Documents disclosed through legal proceedings described the project as an effort to destructively scan books at large scale. Meta has also internally discussed the importance of obtaining large collections of books for AI development.